[PATCH] D152394: [AMDGPU] Switch to backwards scavenging in eliminateFrameIndex

[Jay Foad via Phabricator via llvm-commits]

foad added inline comments.


================
Comment at: llvm/test/CodeGen/AMDGPU/GlobalISel/call-outgoing-stack-args.ll:163-172
+; FLATSCR-NEXT:    scratch_store_dwordx2 off, v[0:1], s0 offset:8
+; FLATSCR-NEXT:    scratch_store_dwordx2 off, v[0:1], s0 offset:16
+; FLATSCR-NEXT:    scratch_store_dwordx2 off, v[0:1], s0 offset:24
+; FLATSCR-NEXT:    scratch_store_dwordx2 off, v[0:1], s0 offset:32
+; FLATSCR-NEXT:    scratch_store_dwordx2 off, v[0:1], s0 offset:40
+; FLATSCR-NEXT:    scratch_store_dwordx2 off, v[0:1], s0 offset:48
+; FLATSCR-NEXT:    scratch_store_dwordx2 off, v[0:1], s0 offset:56
----------------
Good example of resuing s0 instead of lots of different SGPRs.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D152394/new/

https://reviews.llvm.org/D152394